Research On Airport Target Recognition Method Based On Convolutional Neural Network

With the rapid development of remote sensing technology and the rapid improvement of computer computing power,the research on remote sensing images has become a new research hotspot.Remote sensing images contain a large number of targets,among which airports and targets in airports are research objects of special value for modern warfare and civilian services.The target recognition of remote sensing images is to obtain a large-scale remote sensing image from high altitude,and then determine whether the image contains the target of interest,and detect,identify and locate the target of interest.The research direction is in military strategic layout,precision strike,UAV automatic driving and other fields have huge market application prospects.However,due to the complexity of the targets in the remote sensing images,the small scale of the targets,the dense targets,and the influence of the natural environment such as clouds and fog during imaging,the traditional target detection technology is not effective in the target detection and recognition of remote sensing images.When the traditional target detection method is applied to remote sensing image target detection,because the detection process needs to rely on manual judgment of the airport name and aircraft type,the detection speed is slow and the accuracy rate is low,which cannot be applied to modern warfare.This paper takes airports,runways and airplanes as the research objects,and studies feature extraction,feature matching,traditional target detection methods and convolutional neural network technology,improves the detection methods of airports and airplanes,and proposes a new airport runway detection method.The algorithm improves the accuracy of airport target recognition.The main work of the paper is as follows:1.Aiming at the problem that the obtained airport images are susceptible to cloud and fog interference,the classical dark channel dehazing algorithm and the theory of dehazing algorithm based on neural network are firstly analyzed,and then an airport image dehazing algorithm based on GCA-NET is researched and implemented.The smooth expansion technique and the end-toend gated context aggregation network are used to restore the original image,and the gated subnet is used to fuse the features of different levels,which effectively improves the airport image quality.2.In order to improve the accuracy of target detection in airport images,an airport target recognition algorithm based on YOLOv4 was proposed,a data set of relevant airports was constructed,and the training of the airport model was completed to achieve accurate recognition of the target airport.Aiming at the difficulties in aircraft target recognition,such as the changeable attitude and tight arrangement of aircraft targets,an improved YOLOv4 aircraft target recognition algorithm is proposed.First,the input airport image is segmented,then block detection is performed and the detection frames of the overlapping parts are merged,and finally the merged aircraft target detection result is output.The algorithm realizes the accurate identification of different types of aircraft,and improves the accuracy and recall rate of aircraft target detection.3.Aiming at the problem that airport runway extraction is easily affected by the surrounding roads and buildings around the airport,an airport runway detection algorithm based on SURF features and prior knowledge is proposed.Firstly,the location of the target airport in the image to be detected is determined according to the method of feature matching,and then the coordinates of the airport runway are determined according to the fixed prior knowledge of the relative position of the airport runway and the airport.Experiments show that the algorithm can accurately detect the runway in the airport image.4.The software system of airport target recognition is designed and implemented,and each module of the software system is designed according to the requirements of airport target recognition.The design and implementation of user module,image preprocessing module,target recognition module and database management module are introduced in detail.The software system can realize the accurate detection and recognition of airports,airplanes and runways in airport images,which lays the foundation for the practical application of the airport image dehazing algorithm and target recognition algorithm proposed in this paper.